for months I had been running LibreNMS inside a VM including the RRD directory as also the MySQL server.
I didn’t have been cheap with 32GB of RAM, 32 Cores and a decent SSD backed ZFS store.
I had seen contention inside the Linux disc queue therefore I moved the RRD directory out of the VMDK into a dedicated NFS mount which improved the situation a lot but sometimes it still took forever to get graphs showing up primary related to millions of entries withing the syslog and device_perf tables.
Finally I truncated these tables and moved do a dedicated physical MySQL server and now I am really surprised by seeing this change inside all of my graphs.
e.g. this graph has nothing to do with the I/O of my LibreNMS VM, it’s just a quiet router!
Just wonder if someone has an explanation or an idea?